Research And Application Of Blueberry Simulation Software Parallelization Based On GAMA Platform

With the development of high-performance computers,the scale of simulation has been increasing.The application field of simulation has also been extended from the field of military industry to the field of biology,manufacturing,and social economy.Therefore,the parallel simulation technology has become the inevitable choice to accelerate the large-scale simulation system.Based on the model of the wild blueberry insect vector pollination simulation modeling software in the GAMA platform,this thesis deeply studies the distributed real-time parallel simulation technology,and in view of the parallel method of space oriented in the real-time simulation technology,puts forward two key technologies,the partition algorithm of the simulation area and the synchronization algorithm of the simulation.Through the way of simulation based on rectangle region,the blueberry simulation system is divided evenly,for the sake of the solution to the communication problem of parallel blueberry simulation system,through the limited optimistic synchronization algorithm based on Agent aware region.On the basis of the degree of interactivity of different Agent,it is possible to reduce communication traffic in the network by placing the communication-intensive Agent in the same area.Based on the above principles,this thesis uses a simulation space partition method based on rectangular area,selects Agent as a clustering center with larger traffic and larger density and selects a number of Agent,and selects Agent as the next aggregation center according to the threshold value.On this basis,the convex hull algorithm is chosen to make appropriate adjustments to the area in space simulation,and the irregular convex hull is adjusted to the regular rectangular convex hull and the overlapping area is eliminated.This algorithm eliminates redundant traffic caused by overlapping area and irregular boundary.According to the properties and behavior characteristics of the Agent structure,and aiming at the synchronization problem in the space real-time parallel simulation technology,this thesis adopts the limited optimistic synchronization algorithm based on the Agent perception region.The scheduling algorithm,combining the optimism of the optimistic synchronization algorithm and the security of the conservative synchronization algorithm,divides the Agent in the simulation into the secure Agent region and the insecure Agent region,and completes the simulation calculation by the scheduling server and the execution server.The scheduling server selects and determines whether it belongs to the security area according to the security radius of different Agent.The execution server is responsible for executing the behavior of Agent.Through this method,the positive but limited optimistic synchronization in the simulation technology is realized.The algorithm reduces the network communication in the parallel simulation,guarantees the causality constraint in the simulation program running,and shortens the simulation time.